How to Track AEO: Weekly Setup, Metrics, Alerts

To track AEO, choose a fixed set of buyer questions, run them in each AI engine every week, log whether your brand is mentioned and whether your site is cited, and combine that log with referral traffic from AI platforms and your search console data. Alerts then tell you when something important changes, such as losing a citation you used to have.

Our guides on how to measure AEO and AI search visibility and the AEO and GEO KPIs scorecard explain what to measure and why. This post is the setup guide: the exact tracking routine, the sheet to keep, where traffic shows up in analytics and how to automate alerts. It is part of our wider answer engine optimisation resource set.

What does tracking AEO involve?

AEO tracking has three layers, and you need all three because each one answers a different question.

• Visibility: when buyers ask AI engines your key questions, are you mentioned or cited?
• Traffic: do visits arrive from AI platforms, and what do those visitors do?
• Outcomes: do those visits become enquiries, leads and customers?

Visibility without traffic suggests you are mentioned but not linked. Traffic without outcomes suggests the wrong pages are being cited. Tracking only one layer leaves you guessing.

Step 1: Build your prompt panel

A prompt panel is the fixed list of questions you test every week. It is the foundation of everything else, so spend time on it.

• Write 20 to 30 prompts in the words a buyer would actually use.
• Cover every stage: definition ("what is..."), problem ("how do I fix..."), comparison ("how do I choose..."), and selection ("who can help with...").
• Include prompts that name your brand and prompts that do not, because unbranded prompts show whether you win new attention.
• Give each prompt an ID and a topic tag so you can report by theme.

Freeze the panel for at least a quarter. Change it too often and you can no longer compare week to week. Add new prompts at the end rather than editing old ones.

Step 2: Set a consistent testing routine

AI answers vary, so consistency in how you test matters as much as what you test.

• Test on the same day each week and note the date.
• Use a fresh session with no earlier conversation, and the same location and language settings.
• Run each prompt in every engine your buyers use, such as ChatGPT, Google AI Overviews and AI Mode, Perplexity, Gemini and Copilot.
• For important prompts, run twice and record both results, since answers can differ between runs.

Step 3: Log the right fields

Keep one row per prompt, per engine, per week, in a spreadsheet or database. Record these columns:

• Date, prompt ID and engine.
• Brand mentioned (yes or no).
• Domain cited (yes or no) and which URL.
• Position of your citation in the list of sources, if shown.
• Sentiment or framing: positive, neutral, negative or wrong.
• Competing sources cited, as a short list.
• Notes on anything odd, such as outdated facts.

The two numbers to calculate each week are your mention rate (prompts where you were mentioned divided by prompts tested) and your citation rate (prompts where your domain was cited divided by prompts tested), per engine and overall. Here is a tiny script to do that from an exported CSV:

import pandas as pd

log = pd.read_csv("aeo_log.csv", parse_dates=["date"])

weekly = (
    log.groupby(["date", "engine"])
       .agg(tested=("prompt_id", "count"),
            mentioned=("brand_mentioned", "sum"),
            cited=("domain_cited", "sum"))
       .reset_index()
)
weekly["mention_rate"] = weekly["mentioned"] / weekly["tested"]
weekly["citation_rate"] = weekly["cited"] / weekly["tested"]
print(weekly.sort_values(["engine", "date"]).to_string(index=False))

Step 4: Find AI traffic in your analytics

Some AI platforms send visitors with a recognisable referrer or tracking parameter. Common examples include chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com, and links from ChatGPT typically carry a utm_source=chatgpt.com parameter. Platforms change how they send traffic, so check your own referral reports rather than relying on a fixed list.

• Create a channel or segment in your analytics tool that groups these referrers together as "AI assistants".
• Review landing pages for that segment. They show which pages AI engines actually send people to.
• Compare engagement and conversions with other channels.

Be aware that traffic is under-counted. Many people read an AI answer, remember your name and come back later by search or direct visit. For this reason, also track branded search growth and ask new leads how they found you.

Step 5: Use search console data for Google's AI features

Google explains in its AI features documentation how AI Overviews and AI Mode relate to Search Console reporting. Clicks from these features are included in the overall Performance report, though they are not broken out as a separate line in every view. Use the report to watch:

• Impressions and clicks for the question-style queries on your prompt panel.
• Pages that gain impressions without clicks, which may be cited or summarised without a visit.
• Branded query growth, as a sign of recall after exposure.

Bing Webmaster Tools provides similar data and is worth setting up, since Bing's index feeds several other assistants.

Step 6: Set up SERP and citation alerts

A weekly review is the baseline. Alerts catch important changes between reviews. You can combine low-cost tools with your own log.

• Brand mention alerts. Create alerts for your brand name and key product names so you hear when new articles mention you. Google Alerts is a free starting point.
• Threshold alerts on your log. If your citation rate for a key topic drops by a set amount week over week, flag it for review. A simple spreadsheet rule or a scheduled script can send a notification.
• SERP feature monitoring. If you use a rank tracking tool that records AI Overviews and other result features, set alerts for changes on your priority queries. Choose a tool that stores historical snapshots, so you can see what changed and when.
• Accuracy alerts. Add a check for wrong or outdated statements about your company, and assign someone to fix the source whenever one appears.

Keep alerts few and meaningful. Alerts that fire daily get ignored. Define in advance what you will do when each one triggers.

Step 7: Connect visibility to leads

The point of AEO is business growth, so close the loop.

• Tag form submissions and calls from the AI assistants channel in your CRM.
• Add "how did you hear about us?" to forms, with AI assistants as an option plus a free-text field.
• Report on leads and pipeline from these sources alongside the visibility numbers.

Who should own AEO tracking?

Give tracking a single owner, usually someone in marketing or content, with a named backup. The owner runs the weekly test, maintains the log and prepares the monthly report. Content writers, developers and the sales team each get the findings that matter to them: writers see which pages to improve, developers see crawl or rendering problems, and sales sees which questions buyers are asking AI engines. Tracking only pays off when someone is responsible for turning the findings into changes.

A simple monthly report

Summarise on one page, in this order:

1. Headline: mention rate and citation rate this month versus last month, overall and per engine.
2. Wins: prompts where you gained a citation, and the pages responsible.
3. Losses: prompts where you lost one, and a hypothesis why.
4. Traffic and leads: visits and leads from AI assistants.
5. Actions: the three changes you will make next month.

Common AEO tracking mistakes

• Testing a handful of prompts once and drawing conclusions.
• Changing the prompt panel constantly, which breaks comparisons.
• Mixing logged-in, personalised sessions with clean ones.
• Reporting only mentions, ignoring whether links appear.
• Expecting traffic numbers to tell the whole story, when much influence is invisible.
• Collecting data without assigning anyone to act on it.

How Insyrge can help you grow

Insyrge runs AEO tracking as part of its answer engine optimisation service. We build your prompt panel, test it across engines on a fixed schedule, set up AI traffic reporting and alerts, and tie the results to leads in your CRM. You get a clear monthly view of what is working and a prioritised list of what to do next. Start with the free AI visibility check, explore the measurement guide, or contact us to set up tracking for your business. If you are also working on getting cited, see our playbook to get cited by ChatGPT.

Frequently asked questions

How do I track AEO without expensive tools?

Use a fixed prompt panel, a spreadsheet log, your analytics referral reports and Search Console. Add free brand alerts. This is enough to establish a baseline and spot trends before you consider paid tools.

How often should I check AI answers?

Weekly is a good rhythm for most teams. Daily checks mostly capture noise, while monthly checks miss problems for too long.

Is there a SERP alerts tool for AI Overviews?

Some rank tracking tools record whether AI Overviews appear for a query and who is cited, and many can send alerts when that changes. Check that a tool covers the engines and locations you care about and that it keeps history. Pair it with your own prompt log for a fuller picture.

Why does my AI traffic look so small?

Referral data under-counts influence. People often read an answer and later search your name or visit directly. Track branded search, direct traffic trends and lead source answers as well.

Which AEO metric matters most?

Citation rate on your priority prompts is the clearest visibility measure, but only when you connect it to leads. A cited page that never produces an enquiry needs a better next step on the page.

How long before I see changes?

It varies by topic, competition and how much you change. Track for several weeks before judging, and look at trends rather than single weeks.